Earlier this month Canadian news sources, including the CBC and the Toronto Star, reported on Dr Navindra Persaud’s success in securing unpublished data from Health Canada about the safety and effectiveness of Diclectin (a widely used treatment for nausea during pregnancy). To its credit, the country’s drug regulator handed him thousands of pages of data from several clinical trials, including individual patient level data, in order for him to pursue his research. Yet it only did so after Dr Persaud signed a confidentiality agreement. The terms of that agreement provide an important test for Canada’s newly enacted transparency law, since they may limit researchers’ ability to meaningfully provide much needed scientific scrutiny of drug safety and effectiveness.

In late 2014 Canada passed “Vanessa’s Law,” creating several new powers, including the ability to recall drugs from the market and to enforce post-market pharmacovigilance study requirements. After pressure from AllTrials and others, Vanessa’s Law was amended during the legislative process to include a requirement that drug companies make certain “prescribed information” (which has yet to be defined) publicly available. The minister of health was also given the discretion to disclose “confidential business information” to prevent drug related injuries or for the purposes of protecting or promoting human health or public safety.

This new discretionary power in Vanessa’s Law was used to share the Diclectin data with Dr Persaud. But Vanessa’s Law says nothing about the terms under which that sharing can happen. So Health Canada filled in that legal void with this confidentiality agreement. It merits close scrutiny.

Firstly, the agreement is strictly between Dr Persaud and Health Canada. Dr Persaud isn’t allowed to share the data even with a colleague (say, a statistician) to help him complete the proposed analysis.

Secondly, Dr Persaud agreed that the data he received would “not be reproduced in whole or in part in any document, paper, manuscript, etc. that [he] intends to publish or otherwise make public.” He also has to destroy the information after the project is complete. This puts Dr Persaud in a catch 22 situation. The best practice when publishing clinical studies is to make the underlying data publicly available. He needs access to the data to do his research. But in order to obtain the data he has to sign an agreement that prevents him from sharing them. Coupled with an inability to share data in the course of his research, this will significantly hamper scientific exchange of and constructive debate about the interpretation of the data. Responsible scientists appreciate feedback in the process of data analysis, and expect data based challenges after publication.

Thirdly, Dr Persaud is required to “provide Health Canada with a copy of any draft manuscript which was developed using the [data] at least 15 business days prior to submitting it for publication . . . in order for Health Canada to verify that the [data] was used for the purpose of the Project.” This clause doesn’t appear to give Health Canada the authority to veto publications at will. But the clause says nothing about how quickly Health Canada has to turn around study drafts or whether Dr Persaud can go ahead and submit after a reasonable timeframe. This provision could be used to delay publication—which isn’t hard to imagine if Dr Persaud finds, contrary to Health Canada’s approval of the drug, that Diclectin is an ineffective treatment for morning sickness.

How did Health Canada come up with these terms? Our guess is that they looked at the “data use agreements” in place for other data sharing initiatives, such as the Yale Open Data Access (YODA) initiative and the model agreement set by companies participating in ClinicalStudyRequest.com.

There are some broad similarities between these agreements. Like the Persaud-Health Canada agreement, the YODA and model industry agreements restrict use of the data to the specific research project proposed when seeking access to the data. The YODA and industry agreements also preclude researchers from sharing even de-identified data as part of a publication.

But there are salient differences. On the positive side, the YODA agreements require researchers to provide a copy of publications as they are submitted (not 15 days in advance like Health Canada). As well, both the YODA and industry agreements allow for at least some sharing of data for the purpose of carrying out the proposed research (think of the statistician).

On the negative, the YODA and industry agreements require researchers to alert the companies involved (not regulators) of any safety concerns they identify, which could precipitate delayed safety warnings. Data from YODA and ClinicalStudyRequests.com must also be accessed through secure platforms, and not downloaded, which can impede efficient analysis. Finally, it’s clear that researchers have to provide detailed plans for their research, statistical analysis, etc. before they can proceed (especially in the case of ClinicalStudyRequests.com), which has the potential to undermine the level of independent scrutiny that researchers who obtain data access are able to provide.

Even if the Persaud-Health Canada agreement is comparable to the YODA and model industry agreements, the fundamental question is why these agreements should become the standard? Health Canada could have instead looked at the “Terms of Use” (ToU) of the European Medicines Agency’s (EMA’s) new clinical data publication policy. The EMA’s ToU provide access to data for research purposes without any imposition of confidentiality with regards to the data—either during the research process or for the purpose of publication. Researchers have to confirm their identity and the purpose of the research. They also have to agree not to “re-identify” clinical trial participants or use the data for “commercial purposes.” But their use of the data for research, including comparative effectiveness research, is otherwise unfettered.

If the goal of the transparency powers in Vanessa’s Law is to encourage independent scrutiny of drug safety and effectiveness data, thereby contributing to a better protection of the public, Health Canada can and must do better. The focus must be on facilitating Dr Persaud and others like him to analyze the data in order to protect or promote human health and the safety of the public. Instead, Health Canada’s approach appears to make that harder to achieve. The legal warning that the agency will “closely monitor compliance” and take “legal action in the event of the breach”—a point stressed in an accompanying letter to Dr Persaud—can also hardly be seen as encouraging independent scrutiny.

If Health Canada is really committed to transparency, it should exclude drug safety and effectiveness data from the scope of “confidential business information” (which some of us argued for while Vanessa’s Law went through the legislative process). Had it done so already with the data Dr Persaud requested, Canada’s drug regulator could have focused on how to allow publication of the findings and underlying data, while protecting research participants’ privacy through de-identification. With the imposition of its confidentiality agreement, it has imposed excessive secrecy and requested destruction of the data once the research is complete—all under threat of liability.

Matthew Herder, assistant professor, Health Law Institute, Faculties of Medicine and Law, Dalhousie University. Twitter: @cmrherder

Competing interests: MH is a member of the Health Policy Translation research group of the Canadian Center for Vaccinology (CCfV), which has carried out a number of clinical trials sponsored by vaccine manufacturers. MH has not been involved, in any way, with the conduct of these trials, nor received any funds from CCfV to carry out research. He has no other personal, organization, or relational conflicts of interest to disclose.

Trudo Lemmens, professor and scholl chair in health law and policy, Faculties of Law and Public Health and Joint Centre for Bioethics, University of Toronto. Twitter: @TrudoLemmens

Competing interests: None declared.